Preprints
https://doi.org/10.5194/tc-2022-86
https://doi.org/10.5194/tc-2022-86
 
03 May 2022
03 May 2022
Status: this preprint is currently under review for the journal TC.

Sea ice classification of TerraSAR-X ScanSAR images for the MOSAiC expedition incorporating per-class incidence angle dependency of image texture

Wenkai Guo1, Polona Itkin1, Suman Singha2, Anthony Paul Doulgeris1, Malin Johansson1, and Gunnar Spreen3 Wenkai Guo et al.
  • 1Department of Physics and Technology, UiT The Arctic University of Norway
  • 2Maritime Safety and Security Laboratory, Remote Sensing Technology Institute (IMF), German Aerospace Center (DLR), 28359 Bremen, Germany
  • 3Institute of Environmental Physics, University of Bremen

Abstract. In this study, we provide sea ice classification maps of a sub-weekly time series of X-band TerraSAR-X ScanSAR (TSX SC, HH polarization) images from November 2019 to March 2020 covering the Multidisciplinary drifting Observatory for the Study of Arctic Climate (MOSAiC) expedition. This classified time series benefits from the wide spatial coverage and relatively high spatial resolution of the TSX SC dataset, classifying sea ice into leads, young ice with different intensities, and thick ice with different degrees of deformation. We use a classification method considering per-class incidence angle (IA) dependencies (the Gaussian IA classifier) to correct the IA effect (decreasing backscatter with increasing IAs) specific to each class. In addition to HH intensities, we use Gray-Level Co-occurrence Matrix (GLCM) textures as input features to aid the task of one-band classification. Accordingly, we investigate and demonstrate IA dependencies of TSX SC intensities and image textures for different sea ice classes, which are found to be generally lower than those for C-band SAR data. Optimal parameters for GLCM texture calculation are derived to achieve good separation between class distributions while keeping maximum spatial detail and minimizing texture collinearity. Class probabilities yielded from the GIA classifier are further adjusted by a Markov Random Field (MRF) contextual smoothing process to generate final classification results. A significant increase in classification performance is achieved from the inclusion of textures with optimized parameters, as evaluated by classification accuracies (final overall accuracy: 86.05 %) and comparison to sea ice roughness derived from sea ice thickness measurements (correspondence consistently close to or higher than 80 %). Areal fractions of classes representing ice openings (leads and young ice) correspond well with ice opening time series derived from in situ, satellite SAR and optical data in this and previous studies. This study provides a SAR perspective on the changing sea ice conditions surrounding the MOSAiC ice camp through the expedition, and a useful basic dataset for future MOSAiC studies on physical sea ice processes and ocean and climate modeling.

Wenkai Guo et al.

Status: final response (author comments only)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse
  • RC1: 'Comment on tc-2022-86', Anton Korosov, 30 Jun 2022
    • AC1: 'Reply on RC1', wenkai guo, 29 Aug 2022
  • RC2: 'Comment on tc-2022-86', Andreas Stokholm, 21 Jul 2022
    • AC2: 'Reply on RC2', wenkai guo, 29 Aug 2022

Wenkai Guo et al.

Wenkai Guo et al.

Viewed

Total article views: 444 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
320 107 17 444 6 5
  • HTML: 320
  • PDF: 107
  • XML: 17
  • Total: 444
  • BibTeX: 6
  • EndNote: 5
Views and downloads (calculated since 03 May 2022)
Cumulative views and downloads (calculated since 03 May 2022)

Viewed (geographical distribution)

Total article views: 434 (including HTML, PDF, and XML) Thereof 434 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
1
 
 
 
 
Latest update: 20 Sep 2022
Download
Short summary
Sea ice maps are produced to cover the Arctic expedition MOSAiC (2019–2020), and divides sea ice into scientifically meaningful classes. We use a high-resolution X-band synthetic aperture radar dataset, and show how image brightness and texture systematically vary across the images. We use an algorithm that reliably corrects this effect, and achieve good results as evaluated by comparisons to ground observations and other studies. The sea ice maps are useful as a basis for future MOSAiC studies.